How do different exporters react to exchange rate changes? Theory, empirics and aggregate implications

Transcription

1 How do different exporters react to exchange rate changes? Theory, empirics and aggregate implications Nicolas Berman Graduate Institute of International and Development Studies (Geneva) Philippe Martin Sciences Po (Paris) and CEPR Thierry Mayer Sciences Po (Paris), CEPII..and CEPR

2 Motivation Real exchange rate movements are large but small effect on prices and quantities 1) incomplete pass-through of ER movements into import prices: Campa and Golberg (2005), OECD: 50% after 1 quarter; 64% after 1 year Even after conditioning on a price change (Gopinath and Rigobon, 2008): Trade weighted ER pass-through into U.S. import prices is is 22%: Price rigidity not thefullanswer 2) Exchange rate changes have little effect on agregate quantities (exports): Typical macro elasticities are around 1 or just above: much lower than elasticities suggested in trade literature

3 What we do: theory - model with distribution costs, imperfect competition and firm heterogeneity = firms react differently to exchange rate depreciation: - High performance (productivity) firms choose to increase their producer price (increase mark-up) following a depreciation and not their exported volumes. The reverse for low performance firms - Fixed export costs: only high performance firms can export = exporters are firms which, by selection, are more insensitive (in export volumes) to exchange rate movements than other firms = low intensive elasticity: the bulk of exports is concentrated on firms (high performers) that optimally set prices such that their sales do not react much to exchange rates

4 - New firms enter the export market following a depreciation: but they are smaller and less productive than existing ones = low extensive elasticity - Aggregate implications: with sufficient heterogeneity = small effect of exchange rate changes on quantities (as in empirical studies)

5 What we do: empirics Testable implications at the firm level: Very rich firm level data set. Information on firm-level, destination-specific export values and volume from the French Customs and other firm-level information Period to exploit variation across years and destinations. First paper to exploit such a data set to analyze how different exporters react to ER rate movements: we use directly information on the producer prices First paper to report impact of ER movements on entry and exit probabilities

6 Main empirical findings Firms with performance (labor productivity, TFP, export size, number of destinations...) above the median react to a depreciation of 10%: -byincreasingtheirproducerpricebyaround2.0%to3.8%. (othersdonot change their producer price or markup) - by not increasing their volumes (others do increase their sales) - entry probability increases by 1.9% after 10% depreciation; extensive margin 20% of total increase in exports

7 Related literature Distribution costs and the degree of passthrough: Empirics: Campa and Golberg (2007): constitute a share of consumer prices between 40 and 60%; also Burstein, Neves, and Rebelo (2003); Theory: Corsetti and Dedola (2007); Burstein, Eichenbaum, and Rebelo (2005); Closest is Atkeson and Burstein (2008): heterogeneity in market power + trade costs generate deviations from PPP

8 Theory Simple model: Home firms export to countries, one sector (manufacturing) with monopolistic competition standard Dixit-Stiglitz utility: ( )= Z ( ) ( ): consumption of variety : productivity of the firm; 1

9 Transaction costs - iceberg trade cost 1 specific to the pair (Home; country ) - fixed cost to export: - distribution cost: Tirole (1995) production and retailing are complements. Consumer price (in currency ): ( ) ( ) + Distribution costs : any additive cost paid in local currency that does not depend on firm productivity : nominal exchange rate between Home and ( = depreciation vis a vis currency ); ( ): producer price to destination in currency; : wage rate in currency

10 Demand h Demand for a variety: ( ) = ( ) i ( ) = + income; price index in.

11 Optimal prices Producerpricep i (ϕ)inhomecurrencyofvarietyϕsoldincountryi: p i (ϕ)= σ σ 1 ( 1+ η iq i ϕ στ i ) w ϕ } {{ } m i (ϕ) Realexchangerateq i ε iw i w Mark-upm i (ϕ)increaseswithdepreciationandwithproductivity Perceived demand elasticity for producer falls with depreciation and productivity θ i (ϕ)= στ i+η i q i ϕ τ i +η i q i ϕ <σ

12 The impact of a (real) depreciation on the producer price (in domestic currency): ( ) ( ) = + 0 : Endogenous and heterogenous pricing to market Testable Prediction 1.The elasticity of the producer price, ( ) to an increase in is positive and i) increases with the productivity of the firm (and more generally export performance) ii) increases with local distribution costs iii) increases with the level of the real exchange rate

13 Intuition: perceived distribution costs weaken the demand elasticity the more so the lower production costs (the higher the exchange rate and the higher the productivity) Same result if firms differbythequalityofthegood

14 The impact of a change in bilateral RER on the volume of exports: ( ) ( ) = + Testable Prediction 2. The elasticity of the firm exports, ( ) to a real depreciation (an increase in ) is positive and i) decreases with the productivity of the firm ii) decreases with the importance of local distribution costs iii) decreases with the level of the real exchange rate Intuition follows from endogenous pricing to market

15 Profits and the extensive margin Profits for an exporter to increase with depreciation: ( ) = ³ h i 1 ³

16 Threshold prdoductivity of the zero profit firm exporting in : ³ ³ = 1 Only high productivity firms can export: those firms price to market and are less sensitive to RER changes: Selection effect Threshold productivity with depreciation = 1 Entry of less productive and smaller firms triggered by a depreciation

17 Aggregate exports Pareto distribution for productivity: ( ) =1, inverse measure of productivity heterogeneity. ( ) aggregate exports: all individual exports of firms with productivity : = Z ³ h + i ( )

18 Elasticity of aggregate exports to RER = intensive + extensive elasticities: = Z ( ) ( ) {z } ( ) 0 ( ) {z } =

19 Quantitative results Can our model explain the low aggregate(intensive + extensive) elasticities of exports to RER? k=1.5(mayerandottaviano2008);σ=7 Note: noneedtohaverestriction: k>σ asinchaney(2008) Simulate model with τ i =1.2; ϕ i such that P(ϕ<ϕ i )=G(ϕ i )=0.8; 20% of firms export Share of distribution costs in consumer prices: 0.5

20 Results from simulation We can reproduce both low observed intensive (exports from existing exporters) and extensive margins Standard new trade model: elasticity = +0= :toohigh Melitz/Chaney model: elasticity = +( ) = Table 1: Calibration of aggregate export elasticities to exchange rate French data Benchmark =1 =2 =4 =0 3 Intensive Extensive Total

21 Empirics Data: Large database on French firms. 2 sources: 1) French customs for firm-level trade data: export, for each firm, by destinationyear, both in value and volume; 2) Firm-level information from INSEE: sales, employment, sector... Merge the two: virtually all individual French exporters still present (90%) Period: 1995 to 2005

22 Two restrictions: - restrict to single-product firms: proxy of export price = export unit value (ratio of export value/volume per destination): an in export value is an increase in export price only if single product. At product level, no information on productivity (firm level) we lose around half of the observations but less than 10% of exporters (most exporters are single product to at least one destination/year) - restrict to non euro-zone exports exporters left Robustness checks on whole sample and at product level

23 Export unit values. Producer prices (proxied by export unit values): ( ) = depend on 1 ³ 1+ 1), : firm-destination fixed effects 2) : year dummies 3) = :realexchangerate 4) : laggedfirm labor productivity ( )= 1 ( 1 )+ 2 ( )+ + + :average RER between France and during year. Also allow delayed effect on producer prices Testable implication 1: 2 larger for high performance firms

24 TABLE 3 : EXCHANGE RATE AND UNIT VALUES Dep. Var. : Unit Value (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Performance Indicator TFP TFP(t-2) Labor Productivity Nb Destinations Export Volume Sub-sample All High Low High Low High Low High Low High Low TFP(t-1) * * (0.008) (0.013) (0.013) (0.011) (0.013) (0.011) (0.014) Labor Productivity(t-1) (0.013) (0.013) TFP(t-2) (0.020) (0.017) RER 0.166*** 0.212** *** ** *** * (0.056) (0.088) (0.083) (0.102) (0.096) (0.090) (0.080) (0.064) (0.127) (0.071) (0.096) Observations R-squared All variables in logarithms. Robust standard errors in parentheses. Panel, within estimations (firm-destination fixed effects) with year dummies. Subsamples computed by destination-year, except for columns (8) and (9), computed by year. * significant at 10%; ** significant at 5%; *** significant at 1%

25 Export volumes ExportVolumes: x i (ϕ)=y i P σ 1 i [ ] τi σw σ ϕq +η i i i -Y i (GDP),w i (GDPpercapita)andP i (effectiverer) ( ) σ 1 σ dependon Log(x hit )=β 1 Log(ϕ ht 1 )+β 2 Log(RER it )+β 3 Z it +ψ t +µ hi +υ hit Z i : setofdestination-yearspecificvariables(gdp,gdppercapitaandeffective RER). Firm-destinationfixedeffectsµ hi andyeardummiesψ t. Testableimplication2: β 2 lowerforhighperformancefirms

26 TABLE 4 : EXCHANGE RATE AND EXPORT VOLUMES Dep. Var. : Export Volume (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Performance Indicator TFP TFP(t-2) Labor Productivity Nb Destinations Export Volume Sub-sample All High Low High Low High Low High Low High Low TFP(t-1) 0.070*** 0.076** *** 0.094*** (0.020) (0.031) (0.033) (0.030) (0.028) (0.028) (0.030) Labor Productivity(t-1) 0.067** 0.063* TFP(t-2) RER 0.333** *** ** 0.341* 0.566*** *** 0.330* 0.531** (0.130) (0.204) (0.207) (0.258) (0.229) (0.206) (0.204) (0.269) (0.155) (0.176) (0.209) Effective RER *** ** * ** *** * ** (0.081) (0.124) (0.136) (0.151) (0.149) (0.126) (0.131) (0.154) (0.101) (0.110) (0.131) GDP 0.810* *** * *** (0.442) (0.666) (0.748) (0.918) (0.910) (0.666) (0.722) (1.042) (0.531) (0.589) (0.748) GDP per capita ** * (0.677) (0.768) (0.984) (0.955) (0.676) (0.742) (1.132) (0.524) (0.599) (0.763) Observations R-squared All variables in logarithms. Robust standard errors in parentheses. Panel, within estimations (firm-destination fixed effects) with year dummies. Subsamples computed by destination-year, except for columns (8) and (9), computed by year. * significant at 10%; ** significant at 5%; *** significant at 1%

27 TABLE 5 : DISTRIBUTIOM COSTS AND NON LINEAR EFFECT OF EXCHANGE RATE VARIATIONS (1) (2) (3) (4) (5) (6) Dep. Var. Unit Value Export vol. Unit Value Export volume Sub-sample All All High RER Low RER High RER Low RER TFP(t-1) *** *** 0.103** (0.013) (0.033) (0.012) (0.015) (0.029) (0.042) RER * 0.326** ** (0.211) (0.472) (0.128) (0.125) (0.284) (0.351) RER*Distribution 1.910** ** (0.748) (1.625) Observations R-squared Robust standard errors in parentheses. Panel, within estimations (firm-destination fixed effects) with year dummies. Destination-specific controls not reported. Subsamples computed by destination. * significant at 10%; ** significant at 5%; *** significant at 1%

28 TABLE 11 : ROBUSTNESS: PRODUCT LEVEL AND DECILE DECOMPOSITION (1) (2) (3) (4) (5) (6) (7) (8) PRODUCT LEVEL DECILE DECOMPOSITION Dependent Variable Unit Value Export Volume Unit Value Export Volume Performance Indicator: TFP Sub-sample All High Low All High Low 10% High 10% Low 10% High 10% Low TFP(t-1) 0.016*** 0.024*** 0.008** 0.062*** 0.038*** 0.132*** (0.004) (0.004) (0.009) (0.011) (0.011) (0.015) (0.018) (0.037) (0.043) RER 0.157*** 0.205*** 0.110*** 0.407*** 0.312*** 0.489*** 0.227* ** (0.025) (0.026) (0.059) (0.067) (0.069) (0.125) (0.183) (0.304) (0.442) Observations R-squared All variables in logarithms. Robust standard errors in parentheses. Panel, within estimations (firm-destination fixed effects) with year dummies. Destination specific controls not reported. Sub-samples computed by destination-year. * significant at 10%; ** significant at 5%; *** significant at 1%

29 Alternatives - imported inputs whose price increase with depreciation: control for each firm average destination imports - decreasing returns higher MCs with higher sales: control for total sales of the firm - high market power firms price to market: control for share of firms s exports in the country/sector - competition intensity: when we split between high and low productivity firms, we may split between high and low competition sectors: split firms according to the median level of productivity inside each sector

30 TABLE 12 : ROBUSTNESS: ALTERNATIVES (1) (1) (2) (3) (4) (5) (6) (7) (8) ALTERNATIVE IMPORTED INPUTS DECREASING RETURNS Dependent Variable Unit Value Export Volume Unit Value Export Volume Performance Indicator: TFP Sub-sample High Low High Low High Low High Low TFP(t-1) * 0.067* ** 0.024* (0.012) (0.013) (0.030) (0.034) (0.013) (0.013) (0.031) (0.033) RER 0.225** *** 0.211** *** (0.088) (0.083) (0.204) (0.208) (0.088) (0.083) (0.204) (0.207) Imports / Total Sales (0.054) (0.044) (0.105) (0.102) Total Sales 0.054*** *** 0.230*** (0.018) (0.011) (0.040) (0.029) Observations R-squared All variables but Imports/Total Sales in logarithms. Robust standard errors in parentheses. Panel, within estimations (firm-destination with year dummies. Sub-samples computed by destination-year. * significant at 10%; ** significant at 5%; *** significant at 1%

31 TABLE 13 : ROBUSTNESS: ALTERNATIVES (2) (1) (2) (3) (4) (5) (6) (7) (8) ALTERNATIVE MARKET POWER COMPETITION INTENSITY Dependent Variable Unit Value Export Volume Unit Value Export Volume Performance Indicator: TFP Sub-sample High Low High Low High Low High Low TFP(t-1) * 0.066** ** (0.013) (0.013) (0.030) (0.033) (0.013) (0.013) (0.032) (0.033) RER 0.215** *** 0.192** *** (0.088) (0.083) (0.201) (0.205) (0.092) (0.085) (0.210) (0.213) Share of French Exports *** *** (0.288) (0.414) (2.407) (6.303) Observations R-squared All variables but "Share of french exports" in logarithms. Robust standard errors in parentheses. Panel, within estimations (firmwith year dummies. Sub-samples computed sector-destination-year for columns (5) to (8) * significant at 10%; ** 5%; *** 1%

32 Exporting decisions and new exporters - Depreciation should increase probability of exporting: entry + not exiting - Exporters entering due to exchange rate depreciation should be smaller

33 TABLE 6 : EXCHANGE RATE AND EXPORTING DECISIONS (1) (2) (3) (4) (5) (6) (7) (8) (9) Dep. Var. P(X>0) P(X>0) P(X>0) P(X>0) P(X>0) P(X>0) P(X>0) P(X>0) P(X>0) Condition All X(T-1)=0 X(T-1)=1 All X(T-1)=0 X(T-1)=1 All X(T-1)=0 X(T-1)=1 Labor Productivity(t-1) 0.228*** 0.076*** 0.324*** 0.053*** 0.012*** 0.062*** 0.183*** 0.132*** 0.266*** (0.002) (0.003) (0.004) (0.001) (0.003) (0.001) (0.005) (0.007) (0.011) RER 0.898*** 1.258*** 1.154*** 0.199*** 0.180*** 0.244*** 1.582*** 1.186*** 2.009*** (0.033) (0.052) (0.060) (0.007) (0.007) (0.011) (0.045) (0.061) (0.094) GDP *** *** *** *** *** *** 1.501*** (0.113) (0.178) (0.197) (0.026) (0.026) (0.040) (0.157) (0.215) (0.403) GDP per capita 1.648*** 1.234*** ** 0.382*** 0.188*** 0.070* 3.072*** 2.878*** 0.33 (0.112) (0.175) (0.194) (0.025) (0.026) (0.040) (0.154) (0.211) (0.401) Effective RER *** ** 0.465*** (0.021) (0.034) (0.178) (0.005) (0.030) (0.035) (0.029) (0.039) (0.064) Marginal effects (1) Labor productivity(t-1) 0.054*** 0.012*** 0.065*** 0.036*** 0.021*** 0.064*** RER 0.214*** 0.193*** 0.231*** 0.331*** 0.266*** 0.509*** Observations Estimation Probit OLS FE Logit Robust standard errors in parentheses.all estimations include destination fixed effects and year dummies. (1) Marginal effects computed at means. Linear estimations for FE Logit estimations.* significant at 10%; ** significant at 5%; *** significant at 1%

34 Aggregate results High performance exporters do not react significantly to RER changes Aggregate exports should be weakly responsive to RER if exports are concentrated on high performers (high heterogeneity). Aggregate export volumes by sector / destination and estimate reaction to RER: ( )= 1 ( )+ 2 ( 1 ) is a vector of country-specific controls (GDP, GDP per capita and effective RER) + sector destination fixed effects and year fixed effects

35 TABLE 7 : EXCHANGE RATE AND EXPORT VOLUMES, AGGREGATED Dep. Var. : (1) (2) (3) (4) (5) (6) (7) Sectoral Export Volume Sectoral Indicator κ (Pareto Shape) 10% bigger 10% more productive Sub-sample Whole Sample High Low High Low High Low RER 0.903*** 0.753*** 1.133** 0.501** 1.319*** *** (0.218) (0.183) (0.446) (0.215) (0.309) (0.535) (0.240) RER(t-1) ** ** (0.215) (0.211) (0.388) (0.261) (0.293) (0.355) (0.233) GDP 1.469*** 1.505*** 1.345** 1.189*** 1.187*** 1.622*** 1.353*** (0.329) (0.325) (0.630) (0.383) (0.452) (0.558) (0.462) Total effect of RER 1.111*** 1.244*** 0.895* 0.850*** 1.282*** *** (0.290) (0.287) (0.537) (0.292) (0.390) (0.541) (0.376) Observations R-squared Robust standard errors in parentheses.all estimations include sector-destination fixed effects and year dummies. * significant at 10%; ** significant at 5%; *** significant at 1%

36 TABLE 8 : EXCHANGE RATE AND EXPORT VOLUME OF EXISTING EXPORTERS, AGGREGATED Dep. Var. : Sectoral Volume (1) (2) (3) (4) (5) (6) (7) of export, existing exporters Sectoral Indicator κ (Pareto Shape) 10% bigger 10% more productive Sub-sample Whole Sample High Low High Low High Low RER 0.678*** 0.600*** *** ** (0.247) (0.193) (0.525) (0.254) (0.328) (0.544) (0.286) RER(t-1) ** (0.228) (0.216) (0.439) (0.267) (0.278) (0.341) (0.254) GDP 1.691*** 1.590*** 1.789** 1.325*** 1.691*** 2.078*** 1.249*** (0.377) (0.314) (0.784) (0.451) (0.576) (0.712) (0.481) Total effect of RER 0.880*** 0.853*** * 0.962*** *** (0.305) (0.629) (0.311) (0.443) (0.576) (0.391) Observations R-squared Robust standard errors in parentheses.all estimations include sector-destination fixed effects and year dummies. * significant at 10%; ** significant at 5%; *** significant at 1%

37 TABLE 9: EXCHANGE RATE, NUMBER OF EXPORTERS AND MEAN VOLUME OF SHIPMENT (1) (2) (3) Dep. Var Total export volume Number of Exporters Mean Vol. of Shipment RER RER(t-1) GDP 0.903*** 0.544*** 0.359* (0.218) (0.057) (0.213) *** (0.215) (0.043) (0.204) 1.469*** 0.738*** 0.731** (0.329) (0.068) (0.322) Total effect of RER 1.111*** 0.691*** (0.290) (0.059) (0.285) Observations R-squared Robust standard errors in parentheses.all estimations include sector-destination fixed effects and year dummies. * significant at 10%; ** significant at 5%; *** significant at 1%

38 Conclusion - rich description of how firms react to RER changes - high performance firms/exporters are firms that absorb RER changes in their price: their volumes are less sensitive to RER changes - with sufficient heterogeneity, implications for aggregate impact of exchange rate change: - a large portion of exports is due to insensitive firms: weakens intensive margin - firms that enter are small: weakens extensive margin

39 Table 2: Descriptive Statistics Nb. Obs. Nb firms Mean Median 1st quartile 3rd quartile ALL OBSERVATIONS Nb Employees VA / L Number of destinations Number of products by dest SINGLE-PRODUCT OBS. Nb Employees VA / L Number of destinations SINGLE-PRODUCT, NON EURO Nb Employees VA / L Number of destinations

40 TABLE 10 : EXCHANGE RATE AND EXPORT VALUES Dep. Var. : Export value (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Performance Indicator TFP TFP(t-2) Labor Productivity Nb Destinations Export Volume Sub-sample All High Low High Low High Low High Low High Low TFP(t-1) 0.080*** 0.059** 0.078*** 0.055** 0.086*** 0.095*** (0.018) (0.027) (0.030) (0.027) (0.025) (0.026) (0.029) Labor Productivity(t-1) 0.068** 0.084*** (0.029) (0.029) TFP(t-2) (0.043) (0.042) RER 0.462*** 0.332* 0.542*** ** 0.496*** 0.502*** *** 0.308* 0.463** (0.112) (0.175) (0.182) (0.218) (0.210) (0.176) (0.176) (0.234) (0.134) (0.166) (0.183) Effective RER (0.069) (0.104) (0.116) (0.126) (0.129) (0.106) (0.111) (0.132) (0.085) (0.097) (0.114) GDP ** 1.149* ** (0.386) (0.587) (0.654) (0.803) (0.818) (0.591) (0.628) (0.880) (0.464) (0.524) (0.660) GDP per capita * 0.911** (0.394) (0.597) (0.669) (0.853) (0.861) (0.600) (0.644) (0.954) (0.459) (0.536) (0.672) Observations R-squared All variables in logarithms. Robust standard errors in parentheses. Panel, within estimations (firm-destination fixed effects) with year dummies. Subsamples computed by destination-year, except for columns (8) and (9), computed by year. * significant at 10%; ** significant at 5%; *** significant at 1%

41 TABLE 5 : DISTRIBUTIOM COSTS AND NON LINEAR EFFECT OF EXCHANGE RATE VARIATIONS (1) (2) (3) (4) (5) (6) Dep. Var. Unit Value Export vol. Unit Value Export volume Sub-sample All All High RER Low RER High RER Low RER TFP(t-1) *** *** 0.103** (0.013) (0.033) (0.012) (0.015) (0.029) (0.042) RER * 0.326** ** (0.211) (0.472) (0.128) (0.125) (0.284) (0.351) RER*Distribution 1.910** ** (0.748) (1.625) Observations R-squared Robust standard errors in parentheses. Panel, within estimations (firm-destination fixed effects) with year dummies. Destination-specific controls not reported. Subsamples computed by destination. * significant at 10%; ** significant at 5%; *** significant at 1%